您所在位置: 首页>>合作交流>>学术交流>>学术通知>>正文
《高阶动态系统低阶控制以及工业方案》学术交流
2017年09月26日 09:40   审核人:

主题:高阶动态系统低阶控制以及工业方案

主讲人:陈思鲁研究员

主讲人单位:中科院宁波工业技术研究院

邀请人:航天学院刘磊

时间:2017年9月27号,10:00-11:30

地点:航天学院4号楼4楼413会议室

陈思鲁研究员,2000年考入浙江大学国际政治专业攻读法学学位,于2005年和2010年分别在新加坡国立大学获得电气工程学士与博士学位。2011年至2017年,他加入新加坡科技研究局属下的新加坡科技制造研究所(SIMTech),任研究科学家。2017年4月,他被中科院宁波工业技术研究院以“团队人才”计划引进,任先进制造技术研究所精密运动与先进机器人团队研究员。至今他已经在国际期刊和会议上发表论文50余篇。他目前的研究方向是机电系统综合设计和参数优化,以及柔性系统的亚刚体控制。

轻质零部件和柔性连接的低频共振对高精控制提出了新的挑战,而目前工业上应用的控制器仍然是固定结构的线性低阶控制器如PID和惯性前馈,同时现行的运动轨迹规划至多到三阶。因此课题组研发了基于模型的三阶前馈控制器处理刚体模态和柔性模态。对前馈和反馈环融合,并对固定结构控制器基于优化参数,通过机电一体化的设计提出了高效优化算法。

Low-Order Controls for High-Order Dynamics, Where to Go?

Abstract: The physical motion systems may contain multiple resonant modes and be impeded nonlinear disturbances. But for ease of tuning and maintenance, the corresponding industrial motion controllers are generally in low-order, linear, and fixed structures forms, such as the PID and inertia feedforward controllers. Some of our recent proposed approaches to solve above conflict are discussed in this talk. Firstly, to cater for the current industrial practice of the trajectory planning up to the jerk, we have developed a model-based, third-order feedforward controller to concurrently handle both the rigid-body mode and the lump-sum of the flexible modes. Secondly, by incorporating this controller in both feedforward and feedback loops, and performing the follow-up data-based tuning of the fixed structure controller, we are able to achieve better tracking and disturbance rejection without compromising the performance of noise attenuation. Thirdly, to satisfy additional specifications such as energy saving and jerk-decoupling, solely tuning of the fixed structure controller is difficult. If time permits, I will talk about our proposed hybrid system-controller optimization approach, while we find that such a class of problems is not solvable by linear optimal control theory. Here, efficient optimization algorithm is developed to achieve parameter convergence as fast as the convex optimization counterparts.

Close Window